How to Describe Your Audience for an AI Concept Test

On this page

1. Why vague input gives vague results
2. Describe the audience by situation, not by demographics alone
3. Write the idea summary like you’d explain it to a stranger
4. Example: the same idea, vague and specific
5. Check your draft before you spend a test
6. Read the results with the limits in mind
7. Your next step

Say you type “everyone who likes fitness” into an AI concept test. What comes back is as vague as what you put in. The planning step reads your audience and splits it into groups, and “everyone” gives it nothing to split on. So you get bland groups and answers that are polite and don’t say much.

The AI personas only know what you write. That means your audience settings, your idea text, your price and any photo descriptions. They don’t visit your website. They don’t know your competitors. If a detail isn’t in your text, it doesn’t exist for them.

So the real work happens before the test. This article walks through a practical method: name the audience, state the problem, describe the offer in plain words, add the price, then sanity-check the draft before you spend a test on it.

Why vague input gives vague results

Before any participant answers, a planning step runs. An AI model reads the audience settings you typed and your idea, then cuts the crowd into groups. Each group comes back with a name and a share of the audience. It also gets an age range, four typical jobs or life roles, a budget level (tight, average, comfortable or high) and a short note on how those people shop.

Say you write “everyone who likes fitness.” You’ve handed the model nothing, so it guesses. You’ll probably get something close to a census table: young, middle-aged, older. A bit of everything. Not much use to you.

Now say you write “working parents who train at home because the gym doesn’t fit their schedule.” The groups start to look like people. One is tight on budget and only buys on sale. Another is comfortable and will pay for anything that saves time.

Here’s why that matters. Everyone in a group shares the same budget level and the same shopping note. Only age and job change from person to person, plus city in the biggest test size. That’s the catch. The quality of each group comes straight from your description, and a weak note gets copied onto every participant in it.

Your idea text works the same way. If you write it like a brochure, the answers read like one too: “seems interesting,” “could be useful,” “depends on the price.” Polite. Useless. You can’t change a product because someone said “seems interesting.”

Being specific won’t make the answers right. It does give the reasons something to hold onto. Name a price, a user and a problem, and the answers can push back on any of the three. That’s the feedback you can act on.

Describe the audience by situation, not by demographics alone

“Women 25 to 40” tells the planner almost nothing. A nurse on night shifts fits it. So does a stay-at-home parent, and so does a freelance designer. They’d each react to your idea in a different way.

So put the situation in. Say what their day looks like. Say what they already use, or put up with, to deal with this problem. Then spell out what a tight or comfortable budget means for them, because a tight budget for a student isn’t the same as a tight budget for a couple with two incomes and a mortgage.

Working parents, mostly 30 to 45, who train at home in short sessions because the gym doesn’t fit their schedule. They currently follow free videos and skip a lot of weeks. Most watch spending on anything that isn’t a necessity.

That’s three sentences. The planner now has something to build groups from: people short on time, careful with money, already half-solving the problem for free.

Pick one core audience

If your idea could serve lots of different people, you still need to choose one. Test the narrowest group you could realistically reach in your first month, not the big crowd you hope to sell to someday. If you can’t decide which one, How to find your first niche walks through it.

Other audiences can wait for later tests. If you blend them into one description, the vagueness creeps right back in.

Say where they are, if it matters

A city, a type of workplace or a stage of life can change the answer. Location only becomes part of each participant’s profile in the biggest test size (100,000 participants). It still shapes how the planner builds the groups, though, so mention it when it’s part of your real picture.

Write the idea summary like you’d explain it to a stranger

Write it the way you’d explain the idea to someone at a dinner table who has never heard of your business. A few sentences is enough. Cover four things: who it’s for, the problem they have, what you actually offer, and what it costs.

Stick to plain nouns. If you write “a weekly box of pre-chopped ingredients for two,” a participant can picture a box on a doorstep. If you write “a modern meal solution,” there’s nothing to picture, and you’ll mostly get a polite shrug back.

Then leave out the hype and leave in the limits. Say where you deliver. Say what the thing doesn’t do, or how long it takes. Participants judge what you wrote, so if you make the idea sound better than the real thing, the result will be better than the real thing deserves. You’ll feel good about it right up until a real customer shows up.

A booking app for independent hair stylists who rent a chair and keep their own appointment book on paper. It sends clients reminders and lets them rebook by text. It doesn’t take payments. $15 a month, works on phones only.

That’s four sentences, and a skeptic can already push back on the price, the missing payments and the phone-only part. Good. That’s the point.

Match the idea type to the answer you want

Pick the type that fits what you’re really asking. For software, you’re asking whether people would try it. For a physical product, whether they’d buy. For a service, whether they’d consider it. The type sets the three answers each participant gives, so a mismatch gives you a result that answers the wrong question.

Example: the same idea, vague and specific

Imagine a founder who has started a weekly meal-prep delivery. She first describes it like this:

Healthy meals for busy people. Affordable and convenient.

Put yourself in a participant’s seat for a second. Who’s busy? A student, a nurse on night shifts and a dad of three all qualify. “Affordable” means something different to each of them, and “healthy” could be a salad or a protein box. Every group fills the gaps its own way, so answers come back as “sounds nice” or “depends on the price.” Nobody has a reason to say yes or no.

Now the specific version:

Weekly delivery of 5 cooked dinners for two, for working couples in mid-sized cities who skip cooking on weeknights. $70 a week, no contract. Vegetarian and standard menus.

Same business. Very different input.

With that to work from, the planner can build groups around real life. Say, couples with two incomes who order takeout too often, couples watching a tighter budget, and couples who already cook but hate weeknights. Each group gets a budget level and a shopping habit that fits the offer.

Reasons change too. Instead of generalities, you’d expect simulated participants to say things like:

• “$70 a week feels high next to cooking at home”
• “Saves me the 6 pm scramble”
• “Worried five dinners will get repetitive”

Price, time, taste. You can act on every one of those. Test a lower price, add a menu rotation, or rewrite the audience line to reach the couples who say yes.

The vague version can’t give you any of that. It never said enough to be wrong.

Check your draft before you spend a test

Before you spend a test, run your summary through the free Idea Roast. Paste in the audience line and the idea text, exactly as you’d enter them. It takes a minute and costs nothing.

Then look at the open questions it gives back. If they’re about basics, like who this is for, what it costs or what the thing actually does, your description is still too thin. The tool is only reading what you wrote, the same way a participant would. If it’s confused, they will be too.

Remember what the Idea Roast is: one AI’s analysis of a short description. It’s a quick read on your draft, not a verdict on your business.

Read it as a skeptic

Now reread your own text with a cold eye. Could two people picture different products from it? “A tool for small teams” could be a calendar, a chat app or an invoicing add-on. If you can imagine two versions, so can the simulated crowd, and your numbers will blur together.

Look for words that carry no information. “Affordable,” “modern,” “easy,” “high quality.” Swap each one for a fact: a price, a time, a size, a place.

Also check for what’s missing. Is the price in there? Is the delivery area? Is the one thing it doesn’t do? Add them.

Use more than one version when you’re unsure

Sometimes you can’t settle a question by staring at it. Is $70 a week right, or is $60 better? Does “no contract” belong in the first sentence or the last? Write both as separate versions.

All versions go to the same simulated participants in the same requests, so the comparison is like for like inside that one test. That’s much cleaner than running two separate tests and wondering whether the crowd changed. Just keep the versions different in one thing at a time, or you won’t know what moved the result.

Read the results with the limits in mind

When the report lands, remember who answered. AI personas did. Nobody paid anything, and nobody is a real person. What you’re looking at is how a simulated crowd reacted to your wording, which isn’t the same as what happens at launch.

Are those personas more generous than real shoppers? Tougher? You can’t know. Either is possible, and the report won’t say. They also never saw your website or your competitors, so if something isn’t in your description, it doesn’t exist for them.

Treat small gaps as a tie

Say version A shows a little more interest than version B. Don’t crown a winner. Order can nudge the AI, and a rerun gives somewhat different numbers. If the gap is small, call it a tie and decide on other grounds, like which wording is clearer or which price you can actually afford.

A big, consistent gap deserves more attention. Even then, it’s a lead to check, not proof.

Go after the reasons and the weak groups

The percentages matter less than the why behind them. Read the reasons first. Then look at which groups said no, and ask whether they’re people you really want to reach.

In the meal-prep example, you might see couples on a tighter budget balk at $70 a week. That’s a question to put to real couples, not a fact about the market.

So go talk to people. Five to ten conversations will tell you what the simulation can’t: what they do today, what they’d actually pay, and what makes them hesitate in the moment. Every report ends with practical steps for checking results this way, so use them.

Your next step

Open a blank note. Write your audience in two sentences: who they are, what their week looks like, and what they already do about this problem. Then write the idea itself in four. Who it’s for, the problem, what you offer, the price. Leave the limits in, like where you deliver or what it doesn’t do.

Paste both into the free Idea Roast and read the open questions it gives you back. If they’re still about basics, say who it’s for or what it costs, fix the text and run it again. That takes a minute.

Once the draft holds up, you’re ready for a real test. If you’re eligible, your first 1,000-participant test is free at /app?signup=1. Want to know what comes back before you commit? Open the sample report and see how the groups, reasons and next steps are laid out.

Then take whatever looked weak to five or ten real people and ask them. The simulation is a quick first check, nothing more, and their answers are the ones that count.

Written by the SellTo.us Team. Examples in this article are hypothetical. SellTo.us results come from AI-simulated participants, not real people, and don’t predict sales or guarantee success.